Face Generation¶

In this project, you'll use generative adversarial networks to generate new images of faces.

Get the Data¶

You'll be using two datasets in this project:

MNIST
CelebA

Since the celebA dataset is complex and you're doing GANs in a project for the first time, we want you to test your neural network on MNIST before CelebA. Running the GANs on MNIST will allow you to see how well your model trains sooner.

If you're using FloydHub, set data_dir to "/input" and use the FloydHub data ID "R5KrjnANiKVhLWAkpXhNBe".

data_dir = './data'

# FloydHub - Use with data ID "R5KrjnANiKVhLWAkpXhNBe"
#data_dir = '/input'


"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import helper

helper.download_extract('mnist', data_dir)
helper.download_extract('celeba', data_dir)

Found mnist Data
Found celeba Data

Explore the Data¶

MNIST¶

As you're aware, the MNIST dataset contains images of handwritten digits. You can view the first number of examples by changing show_n_images.

show_n_images = 25

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
%matplotlib inline
import os
from glob import glob
from matplotlib import pyplot

mnist_images = helper.get_batch(glob(os.path.join(data_dir, 'mnist/*.jpg'))[:show_n_images], 28, 28, 'L')
pyplot.imshow(helper.images_square_grid(mnist_images, 'L'), cmap='gray')

<matplotlib.image.AxesImage at 0x7fa3340f7780>

CelebA¶

The CelebFaces Attributes Dataset (CelebA) dataset contains over 200,000 celebrity images with annotations. Since you're going to be generating faces, you won't need the annotations. You can view the first number of examples by changing show_n_images.

show_n_images = 25

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
mnist_images = helper.get_batch(glob(os.path.join(data_dir, 'img_align_celeba/*.jpg'))[:show_n_images], 28, 28, 'RGB')
pyplot.imshow(helper.images_square_grid(mnist_images, 'RGB'))

<matplotlib.image.AxesImage at 0x7fa3340a5390>

Preprocess the Data¶

Since the project's main focus is on building the GANs, we'll preprocess the data for you. The values of the MNIST and CelebA dataset will be in the range of -0.5 to 0.5 of 28x28 dimensional images. The CelebA images will be cropped to remove parts of the image that don't include a face, then resized down to 28x28.

The MNIST images are black and white images with a single color channel while the CelebA images have 3 color channels (RGB color channel).

Build the Neural Network¶

You'll build the components necessary to build a GANs by implementing the following functions below:

model_inputs
discriminator
generator
model_loss
model_opt
train

Check the Version of TensorFlow and Access to GPU¶

This will check to make sure you have the correct version of TensorFlow and access to a GPU

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
from distutils.version import LooseVersion
import warnings
import tensorflow as tf

# Check TensorFlow Version
assert LooseVersion(tf.__version__) >= LooseVersion('1.0'), 'Please use TensorFlow version 1.0 or newer.  You are using {}'.format(tf.__version__)
print('TensorFlow Version: {}'.format(tf.__version__))

# Check for a GPU
if not tf.test.gpu_device_name():
    warnings.warn('No GPU found. Please use a GPU to train your neural network.')
else:
    print('Default GPU Device: {}'.format(tf.test.gpu_device_name()))

TensorFlow Version: 1.0.0
Default GPU Device: /gpu:0

Input¶

Implement the model_inputs function to create TF Placeholders for the Neural Network. It should create the following placeholders:

Real input images placeholder with rank 4 using image_width, image_height, and image_channels.
Z input placeholder with rank 2 using z_dim.
Learning rate placeholder with rank 0.

Return the placeholders in the following the tuple (tensor of real input images, tensor of z data)

import problem_unittests as tests

def model_inputs(image_width, image_height, image_channels, z_dim):
    """
    Create the model inputs
    :param image_width: The input image width
    :param image_height: The input image height
    :param image_channels: The number of image channels
    :param z_dim: The dimension of Z
    :return: Tuple of (tensor of real input images, tensor of z data, learning rate)
    """
    inputs_real = tf.placeholder(tf.float32, [None, image_width, image_height, image_channels], name='input_real')
    inputs_z = tf.placeholder(tf.float32, (None, z_dim), name='input_z')
    alpha = tf.placeholder(tf.float32, name='learning_rate')
    
    return inputs_real, inputs_z, alpha


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_inputs(model_inputs)

Tests Passed

Discriminator¶

Implement discriminator to create a discriminator neural network that discriminates on images. This function should be able to reuse the variabes in the neural network. Use tf.variable_scope with a scope name of "discriminator" to allow the variables to be reused. The function should return a tuple of (tensor output of the generator, tensor logits of the generator).

# Added alpha to control Leaky Relu

def discriminator(images, reuse=False, alpha=0.2):
    """
    Create the discriminator network
    :param image: Tensor of input image(s)
    :param reuse: Boolean if the weights should be reused
    :return: Tuple of (tensor output of the discriminator, tensor logits of the discriminator)
    """
    with tf.variable_scope('discriminator', reuse=reuse):        
        # Input layer is 28x28x3
        x1 = tf.layers.conv2d(images, 32, 5, strides=1, 
                              padding='same',
                              kernel_initializer=tf.truncated_normal_initializer(stddev=0.02))
        
        relu1 = tf.maximum(alpha * x1, x1)
        # 28x28x32
        
        x2 = tf.layers.conv2d(relu1, 128, 5, strides=2, 
                              padding='same',
                              kernel_initializer=tf.truncated_normal_initializer(stddev=0.02))
        
        bn2 = tf.layers.batch_normalization(x2, training=True)
        relu2 = tf.maximum(alpha * bn2, bn2)
        # 14x14x128
        
        
        x3 = tf.layers.conv2d(relu2, 512, 5, strides=1, 
                              padding='same',
                              kernel_initializer=tf.truncated_normal_initializer(stddev=0.02))
        
        bn3 = tf.layers.batch_normalization(x3, training=True)
        relu3 = tf.maximum(alpha * bn3, bn3)
        # 14x14x512
        
        # Flatten it
        flat = tf.reshape(relu3, (-1, 7*7*512))
        logits = tf.layers.dense(flat, 1,
                                 kernel_initializer=tf.truncated_normal_initializer(stddev=0.02))
        out = tf.sigmoid(logits)
        
        return out, logits


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_discriminator(discriminator, tf)

Tests Passed

Generator¶

Implement generator to generate an image using z. This function should be able to reuse the variabes in the neural network. Use tf.variable_scope with a scope name of "generator" to allow the variables to be reused. The function should return the generated 28 x 28 x out_channel_dim images.

def generator(z, out_channel_dim, is_train=True, alpha=0.2):
    """
    Create the generator network
    :param z: Input z
    :param out_channel_dim: The number of channels in the output image
    :param is_train: Boolean if generator is being used for training
    :return: The tensor output of the generator
    """
    with tf.variable_scope('generator', reuse = not is_train):
        # First fully connected layer
        x1 = tf.layers.dense(z, 7*7*512, 
                             kernel_initializer=tf.truncated_normal_initializer(stddev=0.02))
        # 25088
        
        # Reshape it to start the convolutional stack
        x1 = tf.reshape(x1, (-1, 7, 7, 512))
        x1 = tf.layers.batch_normalization(x1, training=is_train)
        x1 = tf.maximum(alpha * x1, x1)
        # 7x7x512 now
    
        x2 = tf.layers.conv2d_transpose(x1, 256, 5, strides=2, 
                                        padding='same',
                                        kernel_initializer=tf.truncated_normal_initializer(stddev=0.02))
        
        x2 = tf.layers.batch_normalization(x2, training=is_train)
        x2 = tf.maximum(alpha * x2, x2)
        # 14x14x256 now
        
        x3 = tf.layers.conv2d_transpose(x2, 128, 5, strides=2, 
                                        padding='same',
                                        kernel_initializer=tf.truncated_normal_initializer(stddev=0.02))
        
        x3 = tf.layers.batch_normalization(x3, training=is_train)
        x3 = tf.nn.dropout(x3, 0.9)
        x3 = tf.maximum(alpha * x3, x3)
        # 28x28x128 now
                
        # Output layer
        logits = tf.layers.conv2d_transpose(x3, out_channel_dim, 5, strides=1, 
                                            padding='same',
                                            kernel_initializer=tf.truncated_normal_initializer(stddev=0.02))
        # 28x28xN now
        
        out = tf.tanh(logits)
        
        return out


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_generator(generator, tf)

Tests Passed

Loss¶

Implement model_loss to build the GANs for training and calculate the loss. The function should return a tuple of (discriminator loss, generator loss). Use the following functions you implemented:

discriminator(images, reuse=False)
generator(z, out_channel_dim, is_train=True)

def model_loss(input_real, input_z, out_channel_dim, alpha=0.01):
    """
    Get the loss for the discriminator and generator
    :param input_real: Images from the real dataset
    :param input_z: Z input
    :param out_channel_dim: The number of channels in the output image
    :return: A tuple of (discriminator loss, generator loss)
    """
    g_model = generator(input_z, out_channel_dim, alpha=alpha)
    
    d_model_real, d_logits_real = discriminator(input_real, alpha=alpha)
    d_model_fake, d_logits_fake = discriminator(g_model, reuse=True, alpha=alpha)

    d_loss_real = tf.reduce_mean(
        tf.nn.sigmoid_cross_entropy_with_logits(logits=d_logits_real, labels=tf.ones_like(d_model_real)))
    
    d_loss_fake = tf.reduce_mean(
        tf.nn.sigmoid_cross_entropy_with_logits(logits=d_logits_fake, labels=tf.zeros_like(d_model_fake)))
    
    g_loss = tf.reduce_mean(
        tf.nn.sigmoid_cross_entropy_with_logits(logits=d_logits_fake, labels=tf.ones_like(d_model_fake)))

    d_loss = d_loss_real + d_loss_fake

    return d_loss, g_loss


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_loss(model_loss)

Tests Passed

Optimization¶

Implement model_opt to create the optimization operations for the GANs. Use tf.trainable_variables to get all the trainable variables. Filter the variables with names that are in the discriminator and generator scope names. The function should return a tuple of (discriminator training operation, generator training operation).

def model_opt(d_loss, g_loss, learning_rate, beta1):
    """
    Get optimization operations
    :param d_loss: Discriminator loss Tensor
    :param g_loss: Generator loss Tensor
    :param learning_rate: Learning Rate Placeholder
    :param beta1: The exponential decay rate for the 1st moment in the optimizer
    :return: A tuple of (discriminator training operation, generator training operation)
    """
    # Get weights and bias to update
    t_vars = tf.trainable_variables()
    d_vars = [var for var in t_vars if var.name.startswith('discriminator')]
    g_vars = [var for var in t_vars if var.name.startswith('generator')]

    # Optimize
    with tf.control_dependencies(tf.get_collection(tf.GraphKeys.UPDATE_OPS)):
        d_train_opt = tf.train.AdamOptimizer(learning_rate, beta1=beta1).minimize(d_loss, var_list=d_vars)
        g_train_opt = tf.train.AdamOptimizer(learning_rate, beta1=beta1).minimize(g_loss, var_list=g_vars)

    return d_train_opt, g_train_opt

"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
tests.test_model_opt(model_opt, tf)

Tests Passed

Neural Network Training¶

Show Output¶

Use this function to show the current output of the generator during training. It will help you determine how well the GANs is training.

"""
DON'T MODIFY ANYTHING IN THIS CELL
"""
import numpy as np

def show_generator_output(sess, n_images, input_z, out_channel_dim, image_mode):
    """
    Show example output for the generator
    :param sess: TensorFlow session
    :param n_images: Number of Images to display
    :param input_z: Input Z Tensor
    :param out_channel_dim: The number of channels in the output image
    :param image_mode: The mode to use for images ("RGB" or "L")
    """
    cmap = None if image_mode == 'RGB' else 'gray'
    z_dim = input_z.get_shape().as_list()[-1]
    example_z = np.random.uniform(-1, 1, size=[n_images, z_dim])

    samples = sess.run(
        generator(input_z, out_channel_dim, False),
        feed_dict={input_z: example_z})

    images_grid = helper.images_square_grid(samples, image_mode)
    pyplot.imshow(images_grid, cmap=cmap)
    pyplot.show()

Train¶

Implement train to build and train the GANs. Use the following functions you implemented:

model_inputs(image_width, image_height, image_channels, z_dim)
model_loss(input_real, input_z, out_channel_dim)
model_opt(d_loss, g_loss, learning_rate, beta1)

Use the show_generator_output to show generator output while you train. Running show_generator_output for every batch will drastically increase training time and increase the size of the notebook. It's recommended to print the generator output every 100 batches.

def train(epoch_count, batch_size, z_dim, learning_rate, beta1, get_batches, data_shape, data_image_mode):
    """
    Train the GAN
    :param epoch_count: Number of epochs
    :param batch_size: Batch Size
    :param z_dim: Z dimension
    :param learning_rate: Learning Rate
    :param beta1: The exponential decay rate for the 1st moment in the optimizer
    :param get_batches: Function to get batches
    :param data_shape: Shape of the data
    :param data_image_mode: The image mode to use for images ("RGB" or "L")
    """
    losses = []
    
    # def model_inputs(image_width, image_height, image_channels, z_dim):
    input_real, input_z, lr = model_inputs(data_shape[1], data_shape[2], data_shape[3], z_dim)
        
    # def model_loss(input_real, input_z, out_channel_dim, alpha=0.2):
    d_loss, g_loss = model_loss(input_real, input_z, data_shape[3], alpha=0.1)
        
    # def model_opt(d_loss, g_loss, learning_rate, beta1):
    d_opt, g_opt = model_opt(d_loss, g_loss, learning_rate, beta1)

    
    with tf.Session() as sess:
        sess.run(tf.global_variables_initializer())
        
        for epoch_i in range(epoch_count):
            steps = 0
            for batch_images in get_batches(batch_size):
                steps += 1
                
                # Rescale images to values between -1 -> 1
                batch_images = batch_images * 2
                
                # Sample random noise for G
                batch_z = np.random.uniform(-1, 1, size=(batch_size, z_dim))

                # Run optimizers
                _ = sess.run(d_opt, feed_dict={input_real: batch_images, 
                                               input_z: batch_z,
                                               lr: learning_rate})
                _ = sess.run(g_opt, feed_dict={input_real: batch_images, 
                                               input_z: batch_z,
                                               lr: learning_rate})
                _ = sess.run(g_opt, feed_dict={input_real: batch_images, 
                                               input_z: batch_z,
                                               lr: learning_rate})

                if steps % 10 == 0:
                    # At the end of each epoch, get the losses and print them out
                    train_loss_d = d_loss.eval({input_real: batch_images,
                                               input_z: batch_z,
                                               lr: learning_rate})
                    
                    train_loss_g = g_loss.eval({input_real: batch_images,
                                               input_z: batch_z,
                                               lr: learning_rate})

                    print("Epoch {}/{}...".format(epoch_i + 1, epoch_count),
                          "Discriminator Loss: {:.4f}...".format(train_loss_d),
                          "Generator Loss: {:.4f}".format(train_loss_g))
                    # Save losses to view after training
                    losses.append((train_loss_d, train_loss_g))

                if steps % 100 == 0:
                    show_generator_output(sess, 20, input_z, len(data_image_mode), data_image_mode)


        return losses

MNIST¶

Test your GANs architecture on MNIST. After 2 epochs, the GANs should be able to generate images that look like handwritten digits. Make sure the loss of the generator is lower than the loss of the discriminator or close to 0.

batch_size = 8
z_dim = 100
learning_rate = 0.00005
beta1 = 0.5


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
epochs = 2

mnist_dataset = helper.Dataset('mnist', glob(os.path.join(data_dir, 'mnist/*.jpg')))
with tf.Graph().as_default():
    train(epochs, batch_size, z_dim, learning_rate, beta1, mnist_dataset.get_batches,
          mnist_dataset.shape, mnist_dataset.image_mode)

Epoch 1/2... Discriminator Loss: 2.9049... Generator Loss: 0.1051
Epoch 1/2... Discriminator Loss: 2.3867... Generator Loss: 0.2066
Epoch 1/2... Discriminator Loss: 2.1702... Generator Loss: 0.3162
Epoch 1/2... Discriminator Loss: 1.3390... Generator Loss: 0.5334
Epoch 1/2... Discriminator Loss: 1.1320... Generator Loss: 0.7414
Epoch 1/2... Discriminator Loss: 1.3261... Generator Loss: 0.4537
Epoch 1/2... Discriminator Loss: 1.5066... Generator Loss: 0.4875
Epoch 1/2... Discriminator Loss: 1.5630... Generator Loss: 0.4086
Epoch 1/2... Discriminator Loss: 1.5272... Generator Loss: 0.4886
Epoch 1/2... Discriminator Loss: 1.4768... Generator Loss: 0.4208

Epoch 1/2... Discriminator Loss: 1.5648... Generator Loss: 0.3573
Epoch 1/2... Discriminator Loss: 0.9298... Generator Loss: 0.8278
Epoch 1/2... Discriminator Loss: 1.0386... Generator Loss: 0.7785
Epoch 1/2... Discriminator Loss: 1.2201... Generator Loss: 0.5172
Epoch 1/2... Discriminator Loss: 1.0694... Generator Loss: 0.9992
Epoch 1/2... Discriminator Loss: 1.4954... Generator Loss: 0.4355
Epoch 1/2... Discriminator Loss: 1.3426... Generator Loss: 0.4161
Epoch 1/2... Discriminator Loss: 0.9815... Generator Loss: 0.6991
Epoch 1/2... Discriminator Loss: 1.1623... Generator Loss: 0.5665
Epoch 1/2... Discriminator Loss: 1.2845... Generator Loss: 0.5441

Epoch 1/2... Discriminator Loss: 1.2488... Generator Loss: 0.6935
Epoch 1/2... Discriminator Loss: 1.4413... Generator Loss: 0.4793
Epoch 1/2... Discriminator Loss: 1.4875... Generator Loss: 0.3833
Epoch 1/2... Discriminator Loss: 1.4130... Generator Loss: 1.0880
Epoch 1/2... Discriminator Loss: 1.4718... Generator Loss: 0.5243
Epoch 1/2... Discriminator Loss: 1.2400... Generator Loss: 0.6568
Epoch 1/2... Discriminator Loss: 1.4147... Generator Loss: 0.6779
Epoch 1/2... Discriminator Loss: 1.0393... Generator Loss: 0.8133
Epoch 1/2... Discriminator Loss: 1.4022... Generator Loss: 0.7755
Epoch 1/2... Discriminator Loss: 1.6375... Generator Loss: 0.3569

Epoch 1/2... Discriminator Loss: 1.2564... Generator Loss: 0.6391
Epoch 1/2... Discriminator Loss: 1.2011... Generator Loss: 0.7399
Epoch 1/2... Discriminator Loss: 1.4781... Generator Loss: 0.4557
Epoch 1/2... Discriminator Loss: 1.4917... Generator Loss: 0.3730
Epoch 1/2... Discriminator Loss: 1.3326... Generator Loss: 0.7225
Epoch 1/2... Discriminator Loss: 1.1538... Generator Loss: 0.7746
Epoch 1/2... Discriminator Loss: 1.1517... Generator Loss: 0.9168
Epoch 1/2... Discriminator Loss: 0.9684... Generator Loss: 0.9795
Epoch 1/2... Discriminator Loss: 1.3826... Generator Loss: 0.7188
Epoch 1/2... Discriminator Loss: 1.5233... Generator Loss: 0.4848

Epoch 1/2... Discriminator Loss: 1.3590... Generator Loss: 0.4473
Epoch 1/2... Discriminator Loss: 1.4602... Generator Loss: 0.5265
Epoch 1/2... Discriminator Loss: 1.3490... Generator Loss: 0.6844
Epoch 1/2... Discriminator Loss: 1.0789... Generator Loss: 0.6108
Epoch 1/2... Discriminator Loss: 1.3054... Generator Loss: 0.6503
Epoch 1/2... Discriminator Loss: 1.1633... Generator Loss: 0.8282
Epoch 1/2... Discriminator Loss: 1.2335... Generator Loss: 0.7689
Epoch 1/2... Discriminator Loss: 1.7732... Generator Loss: 0.2735
Epoch 1/2... Discriminator Loss: 1.0852... Generator Loss: 0.9281
Epoch 1/2... Discriminator Loss: 1.0816... Generator Loss: 0.8022

Epoch 1/2... Discriminator Loss: 1.3856... Generator Loss: 0.4286
Epoch 1/2... Discriminator Loss: 1.4517... Generator Loss: 0.4582
Epoch 1/2... Discriminator Loss: 1.0725... Generator Loss: 0.8912
Epoch 1/2... Discriminator Loss: 1.2062... Generator Loss: 0.6378
Epoch 1/2... Discriminator Loss: 1.2990... Generator Loss: 0.5258
Epoch 1/2... Discriminator Loss: 0.9803... Generator Loss: 0.7477
Epoch 1/2... Discriminator Loss: 1.3232... Generator Loss: 0.6030
Epoch 1/2... Discriminator Loss: 1.2539... Generator Loss: 0.7911
Epoch 1/2... Discriminator Loss: 1.6661... Generator Loss: 0.4226
Epoch 1/2... Discriminator Loss: 1.5594... Generator Loss: 0.4226

Epoch 1/2... Discriminator Loss: 1.6820... Generator Loss: 0.3962
Epoch 1/2... Discriminator Loss: 0.8540... Generator Loss: 0.9685
Epoch 1/2... Discriminator Loss: 1.1677... Generator Loss: 0.6245
Epoch 1/2... Discriminator Loss: 1.1185... Generator Loss: 0.7152
Epoch 1/2... Discriminator Loss: 1.1187... Generator Loss: 0.7581
Epoch 1/2... Discriminator Loss: 1.3474... Generator Loss: 0.7028
Epoch 1/2... Discriminator Loss: 1.1916... Generator Loss: 0.5153
Epoch 1/2... Discriminator Loss: 1.0879... Generator Loss: 0.8228
Epoch 1/2... Discriminator Loss: 1.5625... Generator Loss: 0.4168
Epoch 1/2... Discriminator Loss: 1.4814... Generator Loss: 0.5565

Epoch 1/2... Discriminator Loss: 1.4255... Generator Loss: 0.5839
Epoch 1/2... Discriminator Loss: 1.3785... Generator Loss: 0.5012
Epoch 1/2... Discriminator Loss: 1.3827... Generator Loss: 0.7007
Epoch 1/2... Discriminator Loss: 1.8049... Generator Loss: 0.2380
Epoch 1/2... Discriminator Loss: 1.0202... Generator Loss: 0.7471
Epoch 1/2... Discriminator Loss: 1.6428... Generator Loss: 0.3380
Epoch 1/2... Discriminator Loss: 1.8014... Generator Loss: 0.3446
Epoch 1/2... Discriminator Loss: 0.8574... Generator Loss: 0.9297
Epoch 1/2... Discriminator Loss: 1.2813... Generator Loss: 0.5608
Epoch 1/2... Discriminator Loss: 0.9884... Generator Loss: 0.8151

Epoch 1/2... Discriminator Loss: 1.2857... Generator Loss: 0.5861
Epoch 1/2... Discriminator Loss: 1.1225... Generator Loss: 0.5931
Epoch 1/2... Discriminator Loss: 1.1238... Generator Loss: 0.6621
Epoch 1/2... Discriminator Loss: 1.6607... Generator Loss: 0.3919
Epoch 1/2... Discriminator Loss: 1.7506... Generator Loss: 0.4355
Epoch 1/2... Discriminator Loss: 1.3246... Generator Loss: 0.6110
Epoch 1/2... Discriminator Loss: 1.4653... Generator Loss: 0.4785
Epoch 1/2... Discriminator Loss: 1.3191... Generator Loss: 0.6138
Epoch 1/2... Discriminator Loss: 1.4428... Generator Loss: 0.4023
Epoch 1/2... Discriminator Loss: 1.1035... Generator Loss: 0.7189

Epoch 1/2... Discriminator Loss: 1.3164... Generator Loss: 0.8637
Epoch 1/2... Discriminator Loss: 2.0114... Generator Loss: 0.2502
Epoch 1/2... Discriminator Loss: 2.0082... Generator Loss: 0.3138
Epoch 1/2... Discriminator Loss: 1.5635... Generator Loss: 0.4213
Epoch 1/2... Discriminator Loss: 1.4549... Generator Loss: 0.3989
Epoch 1/2... Discriminator Loss: 1.0005... Generator Loss: 0.7406
Epoch 1/2... Discriminator Loss: 1.3026... Generator Loss: 0.5064
Epoch 1/2... Discriminator Loss: 1.1046... Generator Loss: 0.8066
Epoch 1/2... Discriminator Loss: 0.9858... Generator Loss: 0.9076
Epoch 1/2... Discriminator Loss: 1.1864... Generator Loss: 0.6649

CelebA¶

Run your GANs on CelebA. It will take around 20 minutes on the average GPU to run one epoch. You can run the whole epoch or stop when it starts to generate realistic faces.

batch_size = 4
z_dim = 100
learning_rate = 0.000035
beta1 = 0.6


"""
DON'T MODIFY ANYTHING IN THIS CELL THAT IS BELOW THIS LINE
"""
epochs = 1

celeba_dataset = helper.Dataset('celeba', glob(os.path.join(data_dir, 'img_align_celeba/*.jpg')))
with tf.Graph().as_default():
    train(epochs, batch_size, z_dim, learning_rate, beta1, celeba_dataset.get_batches,
          celeba_dataset.shape, celeba_dataset.image_mode)

Epoch 1/1... Discriminator Loss: 3.0796... Generator Loss: 0.1940
Epoch 1/1... Discriminator Loss: 2.4436... Generator Loss: 0.1008
Epoch 1/1... Discriminator Loss: 2.7341... Generator Loss: 0.2029
Epoch 1/1... Discriminator Loss: 2.1775... Generator Loss: 0.1850
Epoch 1/1... Discriminator Loss: 1.4964... Generator Loss: 0.5104
Epoch 1/1... Discriminator Loss: 1.1760... Generator Loss: 0.8411
Epoch 1/1... Discriminator Loss: 1.1624... Generator Loss: 0.5339
Epoch 1/1... Discriminator Loss: 1.3596... Generator Loss: 0.7702
Epoch 1/1... Discriminator Loss: 1.5032... Generator Loss: 0.4170
Epoch 1/1... Discriminator Loss: 0.9318... Generator Loss: 0.7519

Epoch 1/1... Discriminator Loss: 1.3145... Generator Loss: 0.7933
Epoch 1/1... Discriminator Loss: 1.6148... Generator Loss: 0.6306
Epoch 1/1... Discriminator Loss: 0.8534... Generator Loss: 1.2720
Epoch 1/1... Discriminator Loss: 1.5517... Generator Loss: 0.8497
Epoch 1/1... Discriminator Loss: 1.0371... Generator Loss: 0.8824
Epoch 1/1... Discriminator Loss: 1.2483... Generator Loss: 0.8121
Epoch 1/1... Discriminator Loss: 0.9092... Generator Loss: 1.0422
Epoch 1/1... Discriminator Loss: 0.6757... Generator Loss: 1.1508
Epoch 1/1... Discriminator Loss: 1.0849... Generator Loss: 0.8364
Epoch 1/1... Discriminator Loss: 0.5831... Generator Loss: 1.2593

Epoch 1/1... Discriminator Loss: 0.6179... Generator Loss: 1.2145
Epoch 1/1... Discriminator Loss: 0.7717... Generator Loss: 1.1218
Epoch 1/1... Discriminator Loss: 0.9166... Generator Loss: 0.9157
Epoch 1/1... Discriminator Loss: 0.6034... Generator Loss: 1.7943
Epoch 1/1... Discriminator Loss: 0.8853... Generator Loss: 1.0933
Epoch 1/1... Discriminator Loss: 0.8131... Generator Loss: 0.9700
Epoch 1/1... Discriminator Loss: 0.7310... Generator Loss: 1.1700
Epoch 1/1... Discriminator Loss: 1.6156... Generator Loss: 0.6088
Epoch 1/1... Discriminator Loss: 0.6982... Generator Loss: 1.1211
Epoch 1/1... Discriminator Loss: 1.1861... Generator Loss: 0.8813

Epoch 1/1... Discriminator Loss: 0.7984... Generator Loss: 1.0394
Epoch 1/1... Discriminator Loss: 0.6021... Generator Loss: 1.3799
Epoch 1/1... Discriminator Loss: 1.0200... Generator Loss: 0.9055
Epoch 1/1... Discriminator Loss: 0.4613... Generator Loss: 1.5279
Epoch 1/1... Discriminator Loss: 0.8241... Generator Loss: 1.1352
Epoch 1/1... Discriminator Loss: 0.6007... Generator Loss: 1.1992
Epoch 1/1... Discriminator Loss: 0.6910... Generator Loss: 1.3543
Epoch 1/1... Discriminator Loss: 0.6947... Generator Loss: 1.5074
Epoch 1/1... Discriminator Loss: 0.5975... Generator Loss: 1.0723
Epoch 1/1... Discriminator Loss: 0.7751... Generator Loss: 1.4335

Epoch 1/1... Discriminator Loss: 1.0542... Generator Loss: 0.8566
Epoch 1/1... Discriminator Loss: 0.6987... Generator Loss: 1.1761
Epoch 1/1... Discriminator Loss: 0.9213... Generator Loss: 1.0923
Epoch 1/1... Discriminator Loss: 1.1194... Generator Loss: 1.3388
Epoch 1/1... Discriminator Loss: 0.7431... Generator Loss: 0.9680
Epoch 1/1... Discriminator Loss: 0.5320... Generator Loss: 1.4443
Epoch 1/1... Discriminator Loss: 0.2183... Generator Loss: 2.5120
Epoch 1/1... Discriminator Loss: 0.3506... Generator Loss: 1.9246
Epoch 1/1... Discriminator Loss: 0.5339... Generator Loss: 1.3016
Epoch 1/1... Discriminator Loss: 0.3528... Generator Loss: 1.7403

Epoch 1/1... Discriminator Loss: 0.1949... Generator Loss: 2.4026
Epoch 1/1... Discriminator Loss: 0.3105... Generator Loss: 2.0467
Epoch 1/1... Discriminator Loss: 0.1542... Generator Loss: 2.8882
Epoch 1/1... Discriminator Loss: 0.5102... Generator Loss: 1.0802
Epoch 1/1... Discriminator Loss: 0.4266... Generator Loss: 1.6658
Epoch 1/1... Discriminator Loss: 0.4480... Generator Loss: 1.8416
Epoch 1/1... Discriminator Loss: 0.4714... Generator Loss: 1.9363
Epoch 1/1... Discriminator Loss: 0.2872... Generator Loss: 2.0278
Epoch 1/1... Discriminator Loss: 0.7158... Generator Loss: 1.5877
Epoch 1/1... Discriminator Loss: 0.8198... Generator Loss: 0.9128

Epoch 1/1... Discriminator Loss: 0.3577... Generator Loss: 1.8875
Epoch 1/1... Discriminator Loss: 0.6429... Generator Loss: 1.2956
Epoch 1/1... Discriminator Loss: 0.4588... Generator Loss: 2.5975
Epoch 1/1... Discriminator Loss: 0.5568... Generator Loss: 1.4211
Epoch 1/1... Discriminator Loss: 0.6923... Generator Loss: 1.5225
Epoch 1/1... Discriminator Loss: 0.3100... Generator Loss: 2.2112
Epoch 1/1... Discriminator Loss: 0.3708... Generator Loss: 1.7851
Epoch 1/1... Discriminator Loss: 0.3639... Generator Loss: 1.5073
Epoch 1/1... Discriminator Loss: 0.4533... Generator Loss: 1.6626
Epoch 1/1... Discriminator Loss: 0.5113... Generator Loss: 1.7695

Epoch 1/1... Discriminator Loss: 0.3981... Generator Loss: 2.0502
Epoch 1/1... Discriminator Loss: 0.8569... Generator Loss: 1.3013
Epoch 1/1... Discriminator Loss: 0.2921... Generator Loss: 1.6178
Epoch 1/1... Discriminator Loss: 0.3128... Generator Loss: 1.8454
Epoch 1/1... Discriminator Loss: 0.9908... Generator Loss: 0.9472
Epoch 1/1... Discriminator Loss: 0.2830... Generator Loss: 2.3156
Epoch 1/1... Discriminator Loss: 0.6434... Generator Loss: 2.0935
Epoch 1/1... Discriminator Loss: 0.3963... Generator Loss: 1.9614
Epoch 1/1... Discriminator Loss: 1.0386... Generator Loss: 1.2970
Epoch 1/1... Discriminator Loss: 0.5066... Generator Loss: 2.0191

Epoch 1/1... Discriminator Loss: 0.7839... Generator Loss: 1.3098
Epoch 1/1... Discriminator Loss: 0.8191... Generator Loss: 1.3667
Epoch 1/1... Discriminator Loss: 0.3293... Generator Loss: 1.6862
Epoch 1/1... Discriminator Loss: 0.2931... Generator Loss: 2.6744
Epoch 1/1... Discriminator Loss: 0.3402... Generator Loss: 1.8501
Epoch 1/1... Discriminator Loss: 1.1569... Generator Loss: 0.7738
Epoch 1/1... Discriminator Loss: 0.5067... Generator Loss: 1.5107
Epoch 1/1... Discriminator Loss: 0.3447... Generator Loss: 1.7903
Epoch 1/1... Discriminator Loss: 0.7319... Generator Loss: 1.0388
Epoch 1/1... Discriminator Loss: 0.3372... Generator Loss: 2.0932

Epoch 1/1... Discriminator Loss: 0.7421... Generator Loss: 1.0556
Epoch 1/1... Discriminator Loss: 0.5120... Generator Loss: 1.2925
Epoch 1/1... Discriminator Loss: 0.2691... Generator Loss: 2.5536
Epoch 1/1... Discriminator Loss: 0.8441... Generator Loss: 1.2175
Epoch 1/1... Discriminator Loss: 0.6333... Generator Loss: 1.2771
Epoch 1/1... Discriminator Loss: 0.2769... Generator Loss: 1.8315
Epoch 1/1... Discriminator Loss: 0.5320... Generator Loss: 1.7081
Epoch 1/1... Discriminator Loss: 0.3372... Generator Loss: 2.2021
Epoch 1/1... Discriminator Loss: 0.6607... Generator Loss: 1.1200
Epoch 1/1... Discriminator Loss: 0.2864... Generator Loss: 1.7598

Submitting This Project¶

When submitting this project, make sure to run all the cells before saving the notebook. Save the notebook file as "dlnd_face_generation.ipynb" and save it as a HTML file under "File" -> "Download as". Include the "helper.py" and "problem_unittests.py" files in your submission.